Selection of repository

711

Infobox 2: Data repository versus (clinical trial) registry

Registries: A clinical trial registry is a collection of records about clinical trials according to an agreed upon set of metadata (27). In registries accepted by the World Health Organization (WHO) and included in their International Clinical Trials Registry Platform (ICTRP), see 9.1, these records contain a minimum amount of information as defined in the WHO Data Set (25). As of 2019, this data set does not define or require attached artifacts or files. Confusingly, the WHO calls the database behind its Search Portal "Central Repository" (27), when it is in fact a registry.

Data repositories: In contrast, a data repository is a (digital) collection of digital datasets. Although not mandatory, the term nowadays implies a function to make these datasets findable, accessible, and reusable (5) and allows for longer term storage. Technically, a repository consists at least of a backend, a database to store metadata and information, and file server to store the datasets and other digital artifacts, and a web-based frontend that allows users to access the backend.

13.1. Principles

According to the FAIR data principles, research data should be findable, accessible, 713

interoperable, and reusable (4,5), see section 2. Principle F3 mandates that "(meta)data are 714

registered or indexed in a searchable resource" (4). Although the principles do not explicitly 715

mention data repositories, principle F3 implies that research data should be stored in an 716

appropriate repository that follows all principles (5). The European Clinical Research 717

Infrastructure Network (ECRIN) data sharing statement is more explicit and states, that "data 718

and trial documents made available for sharing should be transferred to a suitable data 719

repository" (3) and we support this view. According to the FAIR data principles, research data 720

should be findable, accessible, interoperable, and reusable (4,5), see introduction. Principle F3 721

mandates that "(meta)data are registered or indexed in a searchable resource" (4). Although the 722

principles do not explicitly mention data repositories, principle F3 implies that research data 723

should be stored in an appropriate repository that follows all principles (5). 724

When selecting a repository, clinical researchers therefore should ensure that the repository 725

respects all FAIR data principles as a minimum. Although there are alternative initiatives like 726

CoreTrustSeal (28), the FAIR principles seem to be the most widely accepted. However, other 727

initiatives might evolve over time and become generally agreed standards. Given the lack of 728

generally agreed standards and certification processes, researchers will need to assess the 729

suitability of a repository for their purposes. 730

13.2. Time point

Ideally, the appropriate repository is identified before writing the Data Management Plan (see 732

section 7) and then described therein. We assume that a sponsor/investigator uses the same 733

repository for all her/his projects so this should be feasible. 734

13.3. Identifying potential repositories

So far, no repository exists that is specific for clinical research projects. Therefore, clinical 736

researchers need to identify an appropriate repository by themselves. Many institutions 737

involved in clinical research, like universities, currently maintain their own institutional 738

repository. This might be a good starting point in the evaluation process. Alternatively, 739

universities usually have a central contact point that supports researchers with issues related to 740

data sharing and open science in general (29). 741

For projects that were funded by extramural grants, there might be specific requirements for a 742

repository or even a specific repository mandated. For example, the Bill & Melinda Gates 743

Foundation maintains a list of approved repositories for publications published in Gates Open 744

Research (30). It is also expected that the planned European Open Science Cloud (EOSC) will 745

affect how data from projects funded by the European Union will be shared (31). Repository 746

registries maintain a searchable database of repositories. The largest one is probably r3data, a 747

collaborative project of large European academic institutions. R3data can help locating topic 748

specific repositories, which may be a better choice than an institutional repository because data 749

are more likely to be found in a search for that particular topic. Furthermore, Swiss academic 750

research institutions are currently developing a digital repository for long-term preservation and 751

publishing of research data, Olos (32), to support the publication needs of funders and help 752

researchers to manage research data. 753

Another choice might be Zenodo, which is based at CERN (European Organisation for Nuclear 754

Research). There are also for-profit/commercial repositories such as FigShare and Dryad, 755

although we do not explicitly recommend their use. 756

13.4. Selection criteria

After having identified a set of potential repositories, a researcher will need some explicit 758

criteria to select a repository. We suggest an approach to structure this process which is based 759

on a report by the Digital Curation Centre in Edinburgh (33), shaped as a checklist (Table 10, 760

p. 54). Some items are very specific, others cannot be defined exactly and require adaptations 761

on a project basis and not all aspects might be assessable. 762

Another useful resource are the levels of digital preservation by the National Digital 763

Stewardship Alliance (34). 764

Box 8: Recommendations selection of repository

R21. Select a suitable repository, and include this information in the data management plan. Institutional repositories might be a good choice.

R22. Make data as open as possible, but as closed as necessary (FAIR) 765